Google has released Gemini 4 Argon, a new AI model now available worldwide through the API in AI Studio, the Gemini platform, and for Google AI Ultra subscribers in Antigravity, as well as for Ultra and Pro subscribers in the Gemini app and AI Mode in Search. The model tops the Vals Index, ranking #1 across finance, coding, legal, and tax domains, and is priced at $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%.
Key Takeaways
- Argon is the first model to secure the top spot on the Vals Index, a benchmark covering critical economic sectors.
- Enterprise teams report a 35% rise in agentic code submissions and a 33% reduction in code rollbacks after adopting Argon in Antigravity.
- Open‑source projects such as xTB and the Google ADK Python framework have seen performance gains of up to 2.3× and 40% respectively.
- Argon delivers a 2.85× inference speedup on an A100 GPU for Gemma 4 E2B, boosting token throughput from ~37,900 to over 108,000 tokens per second.
| Metric | Value | Why it matters |
|---|---|---|
| Global Availability | API in AI Studio, Gemini platform, Antigravity, Gemini app, AI Mode in Search | Broad access across developer and enterprise channels |
| Pricing | $2/million input tokens; $10/million output tokens; 95% off cached input | Competitive cost structure for high‑volume workloads |
| Vals Index Rank | #1 | Leadership across finance, coding, legal, and tax tasks |
| Code Submission Increase | 35% | Higher productivity in software engineering teams |
Availability and Pricing
Gemini 4 Argon is distributed through multiple Google channels. Developers can access it via the API in AI Studio and the Gemini platform, while Google AI Ultra subscribers gain access in Antigravity. Ultra and Pro subscribers can use the model in the Gemini app and AI Mode in Search. The introductory price is $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%.
Benchmark Performance and Vals Index Leadership
Argon’s performance was evaluated on the Vals Index, a benchmark that aggregates results across finance, coding, legal, and tax domains. The model achieved the highest score, securing the #1 ranking. Its depth and precision in multi‑step tasks outperformed other frontier models in these critical economic sectors.
Engineering Use Cases Across Google
Within Google, Argon has been applied to several high‑impact engineering problems. In Antigravity, software engineers report a 35% increase in agentic code submissions and a 33% reduction in code rollbacks. Specific projects include:
- Chrome Feature Refactor: Argon helped the Chrome team separate core webpage handling from a ~6 million‑line legacy folder within Chromium’s 40 million‑line codebase. The model produced multiple clean code changes that completed an architectural refactor across hundreds of files in weeks, enabling live webpages to move fluidly between tabs, side panels, and floating windows without reloading.
- ADK Python Optimization: The Google ADK Python framework received performance improvements from Argon. By providing the .agents/skills rulebook and a one‑paragraph goal, Argon autonomously identified optimizations that cut in‑memory session preparation time by 40% for long sessions and sped up cold imports by 21%.
- Quantum Chemistry Package: Researchers used Argon to accelerate the open‑source xTB package. The model achieved up to 2.3× speed in predicting molecular structures and 1.8× speed in conformer search, with changes merged into the official xTB GitHub repository.
- Gemma 4 E2B Inference Speedup: Argon enabled a 2.85× end‑to‑end inference speedup on an A100 GPU via vLLM, increasing throughput from ~37,900 to over 108,000 tokens per second without compromising token parity or quality.
- 3D Visualizations: Using slash commands such as /goal and custom skill prompts in Antigravity, Googlers prototype complex 3D visualizations, high‑fidelity games, and interactive animations that render directly in the browser.
Transforming Business and Professional Workflows
Argon combines advanced reasoning, multimodal perception, expanded domain knowledge, and improved writing capabilities. It delivers clearer communication, adheres to style instructions more faithfully, and reduces hedging and repetition compared to previous Gemini models. Enterprise customers and startups in the early‑access program report strong wins, citing Argon’s ability to handle deep multimodal document comprehension, longitudinal data analysis, and interactive analytical sandboxes.
Cybersecurity and Research Impact
Security specialists and researchers employ Argon to audit critical codebases, test hypotheses, and detect zero‑day vulnerabilities. The model is expanding to hundreds of Fairwind Program partners. Wiz, for example, uses Argon in its “Scan for Good” initiative to protect critical infrastructure. Early access has demonstrated Argon’s real‑world impact across high‑stakes sectors, including government data protection.
Why it Matters
Gemini 4 Argon represents a significant step forward in AI model capability, combining top benchmark performance with practical engineering gains. Its availability across multiple Google platforms lowers the barrier for enterprises to adopt cutting‑edge AI, while its pricing structure supports high‑volume use cases. The documented performance improvements in code optimization, scientific computation, and inference speed illustrate tangible benefits that can translate into faster product releases, reduced operational costs, and enhanced security postures.
Frequently Asked Questions
What is Gemini 4 Argon?
Gemini 4 Argon is a new AI model from Google that excels at complex, multi‑step tasks across domains such as coding, finance, legal, cybersecurity, and science. It tops the Vals Index benchmark and is designed for deep work in real‑world scenarios.
Where can I access Gemini 4 Argon?
The model is available globally via the API in AI Studio and the Gemini platform. Google AI Ultra subscribers can use it in Antigravity, while Ultra and Pro subscribers can access it in the Gemini app and AI Mode in Search.
What are the pricing details?
Argon is priced at $2 per million input tokens and $10 per million output tokens. Cached input tokens receive a 95% discount.
How does Argon compare to other models?
On the Vals Index, Argon ranks #1 across finance, coding, legal, and tax domains, outperforming other frontier models in these critical sectors.
Which industries benefit most from Argon?
Industries that rely on complex, multi‑step reasoning—such as software engineering, finance, legal, cybersecurity, quantum computing, and life sciences—stand to gain significant productivity and performance improvements.
Bottom Line
Gemini 4 Argon delivers unmatched performance on the Vals Index and real‑world engineering gains across Google’s own projects. Its broad availability, competitive pricing, and proven speedups in code, scientific computation, and inference make it a compelling choice for enterprises seeking to tackle the most demanding AI workloads.
Reporting based on: original source.
Comments
Post a Comment